gm.text.Gemma3Tokenizer

gm.text.Gemma3Tokenizer#

class gemma.gm.text.Gemma3Tokenizer(
path: str | os.PathLike = 'gs://gemma-data/tokenizers/tokenizer_gemma3.model',
*,
custom_tokens: dict[int,
str] = <factory>,
)[源代码]

基类: gemma.gm.text._tokenizer.Tokenizer

Gemma 3 的分词器。

path: str | os.PathLike = 'gs://gemma-data/tokenizers/tokenizer_gemma3.model'
special_tokens[源代码]

别名: gemma.gm.text._tokenizer._Gemma3SpecialTokens

FORBIDDEN_TOKENS: ClassVar[tuple[int, ...]] = (_Gemma3SpecialTokens.START_OF_IMAGE, _Gemma3SpecialTokens.END_OF_IMAGE)
VERSION: ClassVar[int] = 3